Coupled inviscid-viscous CFD modeling for data center thermal management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computational fluid dynamics (CFD) models for data center thermal management are time-consuming and inefficient, particularly when dealing with heterogeneous technology and complex airflow patterns, leading to insufficient cooling and high energy consumption.
Innovation Solution
A coupled inviscid-viscous solution method (CIVSM) is employed, which divides the data center domain into viscous, inviscid, and interface regions, using Reynolds-averaged Navier-Stokes equations for viscous regions and potential flow equations for inviscid regions, allowing for faster and more accurate simulations by leveraging the speed of inviscid solvers and the physics of viscous regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full Navier-Stokes CFD/HT models are used for data center thermal management, then modeling accuracy is improved, but solution time increases significantly (taking days to arrive at a reasonably accurate solution)
Solution Approach 1:
The patent segments the data center domain into multiple sub-domains (e.g., cold aisles, hot aisles, rack regions) and applies different modeling approaches to each segment. This allows the use of simplified models in regions where high accuracy is less critical while maintaining detailed Navier-Stokes modeling only in regions where thermal precision is most important, thereby reducing overall solution time while preserving accuracy in critical areas.
Solution Approach 2:
The patent applies local quality by using heterogeneous modeling strategies across different spatial regions - full Navier-Stokes equations in regions requiring high thermal accuracy (such as near heat-generating equipment) and simplified algebraic or reduced-order models in regions where approximate solutions suffice. This selective approach maintains measurement precision where needed while dramatically reducing computational time overall.
2Measurement precision
If fine grids are used in NS-CFD calculations to achieve accurate solutions, then modeling precision is improved, but computational time increases (taking days to arrive at a reasonably accurate solution)
Solution Approach 1:
The computational domain is segmented into regions requiring fine grids (near heat sources, thermal boundaries) and regions where coarse grids are sufficient (bulk fluid regions, distant areas). This selective mesh refinement maintains modeling precision in critical zones while reducing the total number of grid cells, thereby improving computational efficiency without sacrificing accuracy where it matters most.
Solution Approach 2:
The patent implements local quality through adaptive grid refinement - applying fine grids only in local regions where thermal gradients are steep or flow physics are complex, while using coarser grids in regions where flow and temperature fields are more uniform. This approach preserves modeling precision in critical areas while dramatically reducing the overall computational burden and improving productivity.
3Measurement precision
If NS-CFD modeling is applied to optimize data center layout and cooling system, then thermal management accuracy is improved, but the time-consuming nature limits its application in optimization processes
Solution Approach 1:
The optimization process is segmented into multiple stages: initial layout exploration using simplified models for rapid evaluation, followed by detailed NS-CFD analysis only for promising candidates. This hierarchical approach maintains thermal management accuracy for final selections while improving optimization efficiency by avoiding computationally expensive simulations for all possible configurations.
Solution Approach 2:
The patent applies preliminary action by performing coarse-grid or simplified-model simulations first to identify promising layout and cooling system configurations. These preliminary results guide subsequent detailed NS-CFD analyses, ensuring that computational resources are focused on evaluating only the most promising options. This staged approach maintains thermal management accuracy for final decisions while dramatically improving optimization efficiency.
Data Source
AI summary
Computational fluid dynamics modeling of a bounded domain is provided which includes solving iteratively a computational fluid dynamics model of the bounded domain using mass flow boundary conditions that are specified. The solving includes automatically adjusting the specified mass flow boundary conditions for at least one iteration of the solving. The automatically adjusting includes mass balancing flows at a boundary of the bounded domain to provide corrected mass flow boundary conditions for the solving. The mass balancing includes applying different corrections to mass in-flow across the domain boundary, compared with mass out-flow across the boundary. The mass balancing may include multiplying mass in-flows across the boundary by (1−CF), and mass out-flows across the boundary by (1+CF), where CF is a determined correction factor, and net mass flow is positive for mass out-flow exiting the domain, and negative for mass in-flow entering the domain.